CO-225

SPATIAL DATA QUALITY AND DECISION MAKING PROCESS

TALHOFER V.(1), HOFMANN A.(1), MAYEROVá .H.(2), KUBíčEK P.(3)

(1) University of Defence, Faculty of Military Technology, Department of Military Geography and

Meteorology, BRNO, CZECH REPUBLIC ; (2) University of Defence, Faculty of Military Technology,

Department of Mathematics and Physics, BRNO, CZECH REPUBLIC ; (3) Masaryk University, Faculty of

Sciences, Geographical Department, Laboratory of Geoinformatics and Cartography, BRNO, CZECH

REPUBLIC

ABSTRACT

The article discusses the relationship between quality of spatial data and decision-making processes

involving the use of geographic information. To support decision-making processes, spatial analyses have

been increasingly used to model the influence of the surrounding landscape and meteorological conditions

on operations in the terrain. The spatial analyses are strongly influenced by the quality of the source data,

and this quality should be considered in the design of these analyses. The article proposes a spatial data

quality assessment and methodology improvement under restrictive conditions, such as time, human

resources and economic aspects. The proposed methodology is verified on the example of the Cross

Country Movement task.

KEYWORDS

Data evaluation, data matching, data quality, decision making support

1. INTRODUCTION: THE DESCRIPTION OF PRESENT SITUATION Command and control systems are used in various branches of the rescue systems, police, medical

emergency services, and in the armed forces. In these systems, the digital geographic data and information

(DGI) are used more and more frequently. DGI is collected from various sources and various technologies are used for its gathering. The

technological processes and their restrictions do not enable us to provide all information in declared stages

and without any time delay. Different obsolescence of the database content is frequent in products,

generally created from large areas of responsibility (e.g. the State’s digital spatial infrastructures). All

these and many other effects cause the positional and thematic properties heterogeneity in the content of

the database. In spite of this situation, data are stored and used in a common spatial database, or used for

various kinds of spatial analyses. The information obtained from the spatial database is applied to a

decision making process, and its quality has an impact on reliability of the final solution.

The goal of this paper is to propose a method focused on the quality of spatial data and information

evaluation, and reliability of the decision making process. The resulting characteristics of data reliability

can be applied to a command and control system (Talhofer, Kubicek, Brazdilova, & Svatonova, 2007).

The proposed methods will contribute to the increasing quality of decision making process.

2. SPATIAL ANALYSES SUITABLE FOR DECISION MAKING PROCESS Spatial analyses are usually developed as a source for a decision making process that contains demanded

information about geographical and meteorological conditions in the area of interest or responsibility,

which can have an important impact on given or planned activities. Fig. 1 illustrates the scheme of a usual procedure of spatial analysis creation.

Fig. 1 Usual process of spatial analysis for the decision-making process

This process has several advantages and disadvantages. The main advantages are:

- relatively simple solution,

- unambiguous result,

- the commander does not have to think about received result and its properties.

The spatial analyses are created from the given spatial data and by the use of a given mathematical model,

so the quality of spatial data and the quality of the mathematical model can significantly affect the result.

Therefore, the spatial analyses have the following essential disadvantages:

- total dependence on spatial data,

- the quality of result is unknown,

- without any supplementary information about the quality of demanded result, the commander can only

make one decision (he has no choice).

In order to overcome the aforementioned disadvantages, we propose to take the spatial data quality

assessment into account.

3. PROPOSED MODEL OF DATA QUALITY AND DECISION-MAKING PROCESS In the area of the Czech Republic, there are many databases available that can be used also for the systems

of command and control. The basic database (Digital Landscape Model 25, DLM25) that is provided by

the Geographical service of the Army of the Czech Republic contains geographic information at the level

corresponding to a topographic map of 1:25,000 scale. The database is regularly updated, however, with

respect to the range of space it is not updated from the whole area of the Czech Republic at the same time.

However, it is possible to obtain the missing or out-of-date information from other state organizations,

which are responsible for the thematic databases creation (e.g. database of roads, water courses, forests

etc.). These databases are updated more often. The stated databases can be used for consequent

specification or addition of information into the basic DLM25 database. Apart from that, there are also

aerial images or satellite scenes, which are also possible to be used for a quick update of DLM25, available. It is always possible to define the concrete territory of military or non-military units’ operation (area of

interest - AOI). Geographic support of the given activity does not usually require all the geographic

information that the basic database contains. According to the character of the activity, it is possible to

define the required information content and its priorities.

The goal of the project we are dealing with is to suggest a procedure of the improvement of geographic

database when solving a concrete task or a set of tasks.

3.1 Model of procedure

The model of procedure assumes that the AOI is known and geographers have defined their Area of

Cartographic Responsibility (AOCR), which can be the same or a bit larger than AOI to support cross-

border activities. The main tasks for geographers are:

- creation of a local database for AOCR from the central database,

- evaluation of the database contents with respect to the given task or given activity,

- modification of the local database for the given task or activity:

- reduction of unnecessary data,

- addition of missing information from external sources;

- if the missing information is not available neither from external nor internal sources, to find out

possibilities or impossibilities of their acquisition with respect to time, price, etc.,

- evaluation of the functionality of the local database for the given task with the help of user aggregation

function F (see the expression ( 1 ) in the next text),

- creation of spatial analyses to support the command and control system of the activity in the alternatives

according to the model in the figure (see Fig. 2),

- decision of the commander of the operation whether the provided analysis alternatives suit him/her or not.

If the provided analyses do not suit the commander and he/she orders the geographic group to specify them,

it is necessary to require the sources (time, personal, financial, etc.) to increase the functionality of the

local database. In case the sources are granted, a differential model of the increase of usability according to

4.3 will be used. The stated model serves mainly to the geographer, so that the geographer is able to – in

given and often restricted conditions – correctly judge the order of importance for the update or addition of

objects of the local database with respect to the task that is being solved.

In more details: Geographer-analyst can prepare several solutions, such as reliable, less reliable; but

according to the previous knowledge about the given area of interest still bearable within a certain level of

risk. If the commander obtains given solutions and additional information about quality properties, it is up

to him what the next steps will be. Either he is satisfied with all the information he has obtained or he

orders geographers to improve the database and prepare a new version of the given spatial analyses. Fig. 2

illustrates this process.

Fig. 2 Process of spatial analyses for decision-making process with the data quality consideration

Two important questions have to be answered if such a solution is accepted:

1. Is it possible to use the risky or less reliable solution when the reliable solution is not possible to use due

to tactical situation in the battlefield or timely restricted area which given solution passes?

2. What measures are necessary to take in order to increase the reliability of spatial analyses solution and

final decision? What expenses will be necessary for it – financial, personal, time etc.?

The third question is no less important – Is it possible to express previous questions in some formalized

systems? The system of spatial data quality evaluation and application of the Value Analyses Theory

(VAT) (Miles, 1989) should help to answer most of the previous questions. In the next paragraph there is a

short description of the suggested system.

3.2 Assumptions of the solution

Precondition for the success of the proposed procedure is a sophisticated system of quality assessment of

spatial databases, a system of evaluation of their utility value and their cost.

4. THE PROCESS OF DATA QUALITY CONSIDERATION Correctness of analysis results considerably depends on the quality of geographic and meteorological data

in use. The recommendations of international organizations dealing with data and information quality

evaluation – e.g. International Organization for Standardization (ISO), Open Geospatial Consortium (OGC)

and Defence Geospatial Information Working Group (DGIWG) – are necessary to be accepted in the

process of quality evaluation.

4.1 Assessment criteria

From DGI quality review, five essential criteria result. Their assessment gives the baseline for relatively

reliable determination of each product utility value (Talhofer, Hoskova, Hofmann, & Kratochvil, 2009):

- Database content expresses mostly compliance of its definition and users’ needs, i.e. concord of the “real

modelled world” and its model represented by objects and phenomena stored in the database.

- Database quality defines the quality of stored data.

- Database timeliness explains how frequently is the entire database or its elements updated.

- Importance of the area is determined by users’ needs, so that it meets the requirements of processed or

supported area range.

- User friendliness. This criterion defines data usability in various software environment types of GIS

nature reflected mostly in compliance to standard principles.

Each of the criteria is mathematically assessable through independent tests and can be described as a

quality parameter. In the next table (Table 1) there is a list of all used criteria.

Table 1 List of criteria for the spatial geodatabase utility value evaluation

4.2 General assessment of spatial data utility

The product (the whole spatial database or its part covering the given AOCR) can be assessed based on the

above mentioned criteria using a suitable aggregation function F (Talhofer, Hoskova, Hofmann, &

Kratochvil, 2009):

( 1 )

The chosen form of the aggregation function concerns also the case when the user gets data of an area

beyond his interest or redundant data; so that their use could seriously affect or even disable the DGI

functions. The weight of each criterion is marked as pi, where i = 1,…,5. The mentioned aggregation

function proves the product status instantly and its utility rate. It is also applicable to experiments to find

the ways to increase product utility at minimum cost increment.

4.3 Individual DGI benefit cost assessment structure

The organisations, such as the Geographic Service of the Army of the Czech Republic or the Czech Office

for Surveying, Mapping, and Cadastre, are usually responsible for DGI databases development

continuously covering all the Czech Republic area or some parts of the world DLM25, Multinational

Geospatial Co-Production Program (MGCP) or Vector Map Level 1 (VMap1) can be mentioned as

examples of such databases in the military.

The DGI are usually developed and maintained by individual parts of the complete database, such as save

units, map sheets etc. Therefore, it is quite a good idea to assess their utility value in the above-described

system within the established storing units introducing individual benefit value. Similarly, the individual

benefit value can be applied for the selected part of master databases from the given area of interest.

When assessing database utility, it is useful to define ideal quality level at first. The ideal level is used as a

comparison standard to express each criterion compliance level. Using the comparison standard the

individual criteria compliance level and consequently aggregate utility may be assessed.

The compliance level of each individual criterion un,s is given as follows:

( 2 )

where:

- ks is for the value of sth criterion compliance,

- ks* is for the level of compliance of sth criterion or its group criterion of the comparison standard.

Than the aggregate individual benefit value (individual functionality – Un) of the nth save unit is defined

by the aggregation function of the same type as ( 1 ). Therefore:

( 3 )

The individual criteria weights are identical with the weights in database utility value calculation.

Particular criteria usually consist of several sub-criteria (see Table 1). The authors took 20 criteria into

their consideration; hence the equation for calculation the aggregate individual utility value is therefore a

function of 20 variables that characterise the levels of compliance for each individual criterion.

Any modification of selected criterion has an impact on the value of Un. Individual variables are

independent one to another, so the derivation of the function can model the changed utility values or

individual utility values.

( 4 )

where i = 1,…, 5, n = 1,…, N, and N is number of all saved units in the database.

Determination of dU value is thus feasible in two ways regarding the desired information structure. When

assessing individual variables effects on the individual functionality value, while the other variables keep

constant values, it is necessary to differentiate U function as follows:

( 5 )

where x is one of the 20 mentioned variables.

In practice, however, such situations may arise that multiple factors may change at the same time, e.g. the

technical quality of database changes in all its parameters—the secondary data derivation methods will

improve location and attribute accuracy and the data integrity will increase, and moreover the data are

stored in a geodatabase accessible to all authorised users. In this database the data are maintained properly

with respect to all topologic, thematic and time relations. In such a case it is suitable to define dU value as

a total differential of all variables describing the modified factors.

4.4 Improved geographic service products utility using value analysis

Database functionality degree is comparable to the cost necessary for provisions—direct used material,

wages, other expenses (HW, SW, amortisation, costs for co-operations, tax and social payments etc.),

research and development cost, overhead cost and others. Functionality and cost imply relative cost

efficiency (RCE) calculated as follows:

( 6 )

where i = 1,…, ℕ.

Similarly to individual utility value, it is possible to consider the impact of particular variables of expenses

Ei on final RCE. The goal is to find such solution as the functionality will be maximised and the expenses

will be minimize.

5. PILOT STUDY

To verify the VAT the task of Cross Country Movement (CCM) was chosen as an example. CCM can be

solved as a common problem or with consideration of certain types of vehicles (the most frequent or the

weakest in the unit, but in case of armed forces usually off road vehicles). The detailed theory of CCM is

in (Rybansky & Vala, 2010).

The solution can offer to the commander not only one possibility, but the variants from which he can

choose according to his intentions and the current situation at the given area.

5.1 Cross Country Movement

The main goal of Cross Country Movement (CCM) theory is to evaluate the impact of geographic

conditions on of a movement of vehicles in terrain. For the purpose of classification and qualification of

geographic factors of CCM, it is necessary to determine:

- particular degrees of CCM

- typology of terrain practicability by kind of military (civilian) vehicles

- geographic factors and features with significant impact on CCM

As a result of the geographic factors impact evaluation we get three known degrees of CCM:

- GO - passable terrain

- SLOW GO - passable terrain with restrictions

- NO GO – impassable terrain

Geographic factors determining CCM and the selection of the access routes are follows:

- gradient of terrain relief and micro relief shapes

- vegetation cover

- soil conditions

- meteorological conditions

- water sheets, water courses

- settlements

- communications

- other natural and manmade objects

The impact of given geographic factor can be evaluated as a coefficient of deceleration ‘Ci’ from the scale

of 0 to 1. The coefficient of deceleration shows the real (simulated) speed of vehicle vj in the landscape in

the confrontation with the maximum speed of given vehicle vmax. The impact of the whole n geographic

factors can be expressed as the formula:

n = 1, …, ℕ. ( 7 )

The main coefficients of deceleration are listed in the next table (see Table 2).

Table 2 Main coefficients of deceleration

Each coefficient consists of several coefficients of 2nd grade. For example C1 is express as:

where:

- C11 is deceleration coefficient by impact of gradient factor,

- C12 is deceleration coefficient by impact of microrelief factor.

The values of deceleration coefficients are counted for given vehicle (its technical properties) from

ascertained properties of geographic objects stored in the spatial geodatabase.

5.1.1 Example

The vehicle can pass terrain step up to hmax high and trench up to wmax width, but the speed on this

passable obstacles is reduced to one half of maximum speed. If the size of obstacle is bigger, the vehicle

velocity is 0. Properties hmax and wmax are given by the technical description of given vehicles and

comparative values are read from spatial geodatabase. In the mathematical formula the condition can be

express:

( 8 )

Using formula ( 7 ) it is possible to create a cost map in which the value of each pixel is the final

(modelled) speed. The cost map can be as a source for the fastest path, reliable path etc. calculation.

The common army vehicle TATRA 815 ARMAX was chosen for a particular vehicle evaluation.

Table 3 The technical characteristics of TATRA 815 ARMAX (Tatra, 2010)

5.2 Spatial database utility value evaluation

The task of CCM solution could require more information that is available in the master database.

Geographer-analyst has to consider which information and in what quality can he obtain from master

database. E.g. for mentioned C12 coefficient it is necessary to select all microrelief obstacles in the area of

interest (road and railway embankments, excavations, terrain steps, trenches etc.). Further he has to find

out all their properties and their accuracy or count how many characteristics are missing. The system

presented in the Table 1 serves as a manual. Next step is the individual functionality value of given part of

master database evaluation.

Two processes were accomplished to verify theory of utility value evaluation. Only data with full

information (all attribute properties had to be filled) were considered in the first case. If some information

was missing in any geographic object system didn’t consider it and one reliable path was analysed. The

first process was in progress according to next schema (Fig. 3):

- creation temporally spatial database for CCM solution from appropriate data and information going from

master database DLM25),

- CCM evaluation - only reliable information are considered,

- final cost map calculation,

- minimum cost path calculation.

Fig. 3 Spatial analyse without database quality evaluation

Only one result is offered to commander. This result seems to be appropriate, but geographer-analyst

generally doesn’t know details about situation on area of responsibility and commander’s intentions.

Problems should appear when the tactical (or other) situation doesn’t make the published path possible to

use. Commander usually requires a new solution in such a case to miss prohibited area. The quality

characteristics of temporally database are than to be considered by geographer to be sure, where are the

weakest points of a new analysis. The weak points of analysis have to be sent to commander together with

own analysis and it is up to him what will be the final decision. Two tasks for commander appear in CCM

example:

1. Use less reliable path and consider that vehicles could stay in front of some obstacle

2. Wait and order to GEO team to improve spatial database (e.g. required properties) as soon as possible

and then use new reliable path

If commander decides to wait for a new solution, geographer has to consider which features are the most

important for given task and tries to improve their quality. The VAT is a suitable tool that can help him to

make an appropriate solution usually under pressure of time and personal restrictions. More detailed

description is available in (Talhofer, Hoskova, Hofmann, & Kratochvil, 2009).

The application of VAT and reliability consideration can be described as follows:

- Creation local spatial database for CCM solution from appropriate data and information going from

master database (DLM25).

- Local spatial database quality evaluation and eventually comparing with the etalon quality (Un = 0.8830

in the example, etalon functionality is F = 1.0068).

- CCM evaluation – only obstacles with fully added properties are considered as obstacles and their

properties are compared with the technical parameters of given vehicle. The other obstacles are considered

only as potential obstacles.

- Final cost map calculation.

- Minimum cost path (Path 1 in the picture Fig. 5) and reliable path (Path 2) calculation. Minimum cost

path assumes all obstacles without properties are probably passable. Reliable path is the same as in the first

case.

In case of mentioned task of CCM, the percentage of filled properties of obstacles were increased of 65%

in comparison with the previous stage and then the utility value of local database increased to 0.9132.

Improved local database served as a new source for the cost map calculation and new reliable path (Path 3

in Fig. 5) calculation.

Fig. 4 Spatial analyse with database quality evaluation and alternative results creation

Fig. 5 Different results of CCM

6. FUTURE WORK - CCM MODELLING WITH UNCERTAINTY All the above described geographic factors (topography, vegetation, soils, weather, hydrology, build-up

area, road network) have been so far treated as certain information with crisp boundaries and varying

levels of (attribute) quality. Spatial categories were treated as crisp sets, with each individual location

exactly identified within a single category. However, the process of geographical abstraction and

generalization about the real world phenomena imposes a vast degree of approximation and uncertainty

(Zhang & Goodchild, 2002). In order to overcome this situation, we can employ two basic approaches –

probabilistic or fuzzy sets theories. For the chosen background data model, the fuzzy set approach seems

to be more appropriate. Fuzzy approach was postulated by (Zadeh, 1965) and further developed by

(Burrogh, Frank, & eds., 1996) within the field of spatial data modelling. It assumes that the traditional

Boolean set theory, in which the assignments are crisp, fails to represent uncertain (vague) entities or

categories. Under fuzzy set theory, the transition between membership and non-membership is gradual,

and any location (pixel in case of raster representation) belongs to fuzzily defined classes valued within the

unit interval <0, 1>. Further, fuzzy membership values for a location typically sum to 1.0 across all classes

(Zhang & Goodchild, 2002).

Only soils and soil cover was chosen for fuzzyfication at the first step. The reason of choosing soils was

twofold:

- the origin of soil map data—soil polygon boundaries are defined based on field observation and

compared with human interpretation, and soil types are classed according to soil scientists expertise. For

many reasons, one can say that soil map is one of the most imprecise maps in the world (Sunila &

Hottanainen, 2004).

- high level of theoretical studies using fuzzy methods for soil mapping and research (Křemenová, 2004),

(Zhu, Burt, Moore, Smith, Liu, & Qi, 2007).

The core idea of this research is to apply fuzzy modelling to the management of expert knowledge in soil

mapping.

7. DISCUSSION AND CONCLUSION It is obvious; the spatial data quality has the significant impact on spatial analyses. The authors tried to

present a relationship between spatial data, type of task which has to be solved, and the quality and the

reliability of the final result. In order to present the whole system alternative Cross Country Movement

modelling for vehicle Tatra 815 has been done – one without the spatial database quality evaluation and

the other using the quality evaluation methodology. Using the quality evaluation procedure the commander

gets more reliable path with known parameters of its quality and also alternative paths with respect to

operational priorities.

Thus the proposed system can complement the existing systems evaluating the costs and technical quality

with the system evaluating user requirements. In addition to that, it provides geographers in the field with a

tool for prioritizing buildings that should be up-dated during rapid update of the local database.

In order to further development of ideas presented above we consider to involve uncertainty measures

within the solution. The core idea is to start with uncertainty of just one spatial component and to apply

fuzzy modelling to the management of expert knowledge in soil data sources. Fuzzy soil maps are then

used in map overlay and CCM evaluation. The goal is to construct a fuzzy soil map presenting

indeterminate soil polygon boundaries. The map will be created in a scale (1:200K) for a CCM calculation

and we believe that fuzzy soil layer with imprecision will be a better input to the analysis than artificial

crisp polygon map with no information about the uncertainty of the boundaries. According to

(Horttanainen & Virrantaus, 2004) modelling uncertainty via fuzzy set theory should also raise the

question how the results of the uncertainty analysis are interpreted by the users who need answers to their

questions about the risk in their decision making. In a military terrain analysis application the related

decision is whether “to go or not to go”.

8. ACKNOWLEDGEMENTS Research results presented above were kindly supported by the project “The evaluation of integrated

digital spatial data reliability” founded by the Czech Science Foundation and also by the Project of

Defence Research of the Ministry of Defence of the Czech Republic “Geographic and meteorological

factors on the theatre”, project code METEOR. Partial results are components of the research task

“Dynamic geovisualization in crises management” supported by the Czech Ministry of Education (grant

MSM0021622418).

9. BIBLIOGRAPHY Burrogh, P. A., Frank, A. U., & eds. (1996). Geographic Objects with Interminate Bounaries. Bristol:

Taylor & Francis Inc. Horttanainen, P., & Virrantaus, K. (2004). Uncertainty Evaluation of Military Terrain Analysis by

Simulation and Visualization. Geoinformatics 2004, Proc. 12th Int. Conference on Geoinformatics -

Geospatial Inforation Research (p. 8). Gävle: Pacific and Atlantic University of Gävle.

Křemenová, O. (2004). Fuzzy Modeling of Soil Maps. Helsinky: University of Technology.

Miles, L. D. (1989). Techniques Of Value Analysis Engeneering (3rd ed.). USA: Eleanor Miles Walker.

Rybansky, M., & Vala, M. (2010). RELIEF IMPACT ON TRANSPORT. ICMT'09: INTERNATIONAL

CONFERENCE ON MILITARY TECHNOLOGIES (pp. 551-559). Brno: UNIV DEFENCE, BRNO.

Sunila, R., & Hottanainen, P. (2004). Fuzzy Model for Soil Polygons for Mapping the Imprecision.

Proceedings of the StartGIS 2003, (p. 8). Pörtschach.

Talhofer, V., Hoskova, S., Hofmann, A., & Kratochvil, V. (2009). The system of the evaluation of

integrated digital spatial data realibility. 6th Conference on Mathematics and Physics at Technical

Universities (pp. 281-288). Brno: University of Defence.

Talhofer, V., Kubicek, P., Brazdilova, J., & Svatonova, H. (2007). DYNAMIC CARTOGRAPHIC

VISUALISATION IN A PROCESS OF TRANSPORTATION MONNITORING OF DANGEROUS

CHEMICAL SUBSTANCES. ICMT '07: INTERNATIONAL CONFERENCE ON MILITARY

TECHNOLOGIES (pp. 597-607). Brno: UNIV DEFENCE, BRNO.

Tatra, a. (2010). Tatra is the solution. Retrieved 5 17, 2010, from TATRA:

http://partners.tatra.cz/exter_pr/vp/new/typovy_listprospekt.asp?kod=341&jazyk=CZ

Zadeh, I. (1965). Fuzy Sets. Information and Control , 8, pp. 338 - 353.

Zhang, J., & Goodchild, M. (2002). Uncertainty in Geographical Information. London: Taylor and Francis,

Inc.

Zhu, A. X., Burt, J. E., Moore, A. C., Smith, M. P., Liu, J., & Qi, F. (2007). SoLIM: A new Technology

For Soil Mapping Unsing GIS, Expert Knowledge & Fuzzy Logic. Overview of SoLIM Programme.

University of Wisconsin.